The short version
- Savings is the top money question for AI: in a J.D. Power survey reported by the ABA (opens in a new tab), savings strategies were the most popular money topic people asked AI about (45%), and 59% of respondents used AI for banking and financial services occasionally.
- The rate gap drives the search: the FDIC’s (opens in a new tab) national average savings rate was 0.38% in June 2026, while Ally reported an average retail deposit yield of 3.12%.
- Depositors are worth a lot: Ally holds $144 billion of retail deposits from 3.6 million deposit customers, and SoFi (opens in a new tab) says its deposits cost 156 basis points less than its warehouse funding, worth about $712.6 million a year.
- Savers move: in a Credit One Bank survey (opens in a new tab) of 1,000 US savers in April 2026, 48% had moved savings to another bank in the previous 12 months.
- AI gets rates wrong: when Which? (opens in a new tab) tested ChatGPT in the UK, it quoted Marcus at 4.5% when the account paid 3.75%, and gave an out-of-date deposit protection limit.
This guide covers how digital banks are found and described in AI answers. It is not financial, legal or regulatory advice; check rate, fee and deposit insurance wording with your compliance team.
Who chooses a digital bank, and what is a depositor worth?
Rate-aware savers, younger adults and people at life events choose them; each depositor brings years of low-cost funding.
By digital bank we mean a chartered bank that serves customers mainly online: Ally Bank, SoFi Bank, Marcus by Goldman Sachs, Charles Schwab Bank and the online savings brands of large banks. App brands that rely on a partner bank are a different case, covered in our neighboring guide on neobanks.
Who switches tells you who to reach. In the Credit One survey, roughly two-thirds of Gen Z respondents had moved some savings to another bank for a higher rate, against fewer than 3 in 10 baby boomers. Expecting parents were the most active rate shoppers: nearly 7 in 10 had moved savings in the past year. At Ally, millennials and younger customers remain the largest generation among new customers.
The value of a depositor shows in the banks’ results:
| Bank | Latest public figures | Why deposits matter |
|---|---|---|
| Ally | $144 billion of retail deposits, 92% FDIC insured; 3.6 million deposit customers; 63 thousand net new in the second quarter of 2026 | Deposits made up 87% of Ally’s funding |
| SoFi | $45.5 billion of deposits, up $5.3 billion in the quarter; 15.8 million members | Deposits replace costlier credit lines, and existing members opened 51% of new products |
The revenue path, we infer, is two-step: a depositor first lowers the bank’s funding cost, then becomes a customer for loans, cards or investing. SoFi says it aims to acquire members through one product and then sell them others, “potentially lowering customer acquisition costs and increasing lifetime value.” Banks that also offer investing can see how investing apps get chosen in AI answers.
Where does AI enter a saver’s decision?
Early, and often: savings questions lead the money topics people take to AI assistants.
- United States. J.D. Power found that thirteen percent of respondents use AI for banking and financial services daily, and another 59% use it occasionally. ChatGPT was the most popular tool, particularly for people under 40; Google Gemini was preferred by people over 40.
- United Kingdom, as a signal. Which? cites a Lloyds Bank study finding 53% of UK adults use AI tools to help with savings. In STRAT7’s April 2026 research (opens in a new tab), 55% of UK adults use AI platforms for financial questions.
Trust is the gap. In STRAT7’s study, ChatGPT scored 53% on trust for financial information, against 92% for bank websites. A reasonable expectation is that savers use AI to build a shortlist and explain options, then go to the bank’s own site to confirm the rate and the protection before they open an account. That makes the bank’s own pages part of the AI journey, not separate from it.
Many savers have already crossed the online line. In a 2024 Bankrate survey reported by InvestmentNews (opens in a new tab), 51 percent of Americans had a savings or money market account with an online bank. Those who did not use one cited a preference for local branches (45 percent), satisfaction with their current institution (42 percent) and concerns over the security of their money (32 percent). Each of those objections is a question a saver can now put to an assistant.
Which questions do savers ask AI before opening an account?
Questions about the best rate, the conditions behind it, safety and fees. We wrote the examples below to illustrate; they are not observed prompts.
| Need | Illustrative prompt |
|---|---|
| Best rate | “Which online savings accounts pay the highest rate right now with no minimum balance?” |
| Comparison | “Ally vs Marcus vs SoFi for an emergency fund” |
| Lump sum | “Where should I put $25,000 I won’t need for a year: savings account or CD?” |
| Safety | “Is my money safe in an online-only bank, and is it FDIC insured?” |
| Conditions | “Does this bank’s savings rate include a temporary bonus, and what happens after it ends?” |
| Fees | “Online checking accounts with no monthly fees and early direct deposit” |
| Switching | “My bank cut my savings rate. Is it worth moving my money?” |
The switching prompt matters most. In the Credit One survey, if a bank cut its rate by a full percentage point, fewer than 2 in 10 savers would switch immediately, but two-thirds would use the cut as a trigger to shop around. Just over half said eliminating monthly fees would prompt them to switch even at the same rate. Each rate move, we infer, sends a wave of savers back to search and AI to compare.
How does a saver get from an AI answer to a funded deposit account?
Through a short chain that breaks easily: named, checked, opened, funded, kept.
- Named. A saver asks for the best options and the answer lists a few banks, usually with rates.
- Checked. The saver confirms the rate, the conditions and deposit insurance, often on the bank’s own site.
- Opened. The account is opened online in minutes.
- Funded. Money moves in. This is the step that creates value for the bank.
- Kept. The depositor stays until a better offer appears, and may add loans or investing later.
An AI answer can break the chain at step two. If it quotes a rate higher than the one on offer, the saver arrives expecting a deal the bank does not offer, and may leave disappointed. If it quotes a rate that is too low or out of date, the bank may never be considered. In both cases the bank sees nothing in its analytics. We suggest tracking AI visibility against funded accounts, not clicks, and asking new customers during account opening where they first heard of the bank. The measurement problem is covered in what to measure in AI visibility.
Why do rates and deposit facts go wrong in AI answers?
Because rates change faster than many pages that AI answers draw on, and conditions are easy to drop.
Observed by a consumer group. On 10 March 2026, Which? asked the free version of ChatGPT about UK savings and checked the answers against Moneyfacts data. Most average and top rates were wrong; only regular saver rates were right. ChatGPT said Marcus offered 4.5%, when its deal paid 3.75%. It suggested a Chase account at 4.5% without saying the rate drops to 2.25% once a temporary bonus ends. It also gave the old UK deposit protection limit of £85,000, though the limit had risen to £120,000 on 1 December 2025, and answers varied when the question was repeated. OpenAI told Which? that for consumer products it recommends using ChatGPT’s built-in search tool, which shows sources. This was one journalist’s test in another country, but the failure modes apply to any bank that competes on rate.
Observed in our studies.
- Prices are often quoted with a twist. In our pricing accuracy study of software plans, 61.9% of plan prices in AI answers were fully faithful to the vendor’s page and 7.6% differed for the same plan and term. Most differing figures came from another page on the vendor’s own site. A savings rate is a price, and the same risk applies, we infer, to old promotional pages.
- Assistants lean to newer pages. In our freshness study, pages published in the last 90 days made up 17.4% to 22.6% of each assistant’s dated citations, against 6.9% of Google’s top 10.
- Finance answers depend on the country. In our country study, on questions such as insurance, lending and tax, the gap between same-country and different-country brand overlap was 0.180, against 0.034 for global products such as headphones. US savers should get US banks; that only works when a bank’s pages make its market and regulator plain.
What decides whether a digital bank is named?
Independent evidence and clear, current facts; the platforms document how they search, not how they pick banks.
Documented. Google says AI Overviews and AI Mode may use “query fan-out” (opens in a new tab), issuing multiple related searches across subtopics. No assistant publishes how it ranks banks.
Observed in our studies. In our frequency study, 88.0% of US financial services and insurance keywords triggered an AI Overview. Of everything our brand entity study measured, coverage on independent sites predicted recommendations best: when the number of independent sites naming a brand in the cited pages rose tenfold, the brand had 4.7 times the odds of being recommended.
Independent evidence banks already earn. J.D. Power’s 2026 Direct Banking Satisfaction Study (opens in a new tab) rated online bank high-yield savings at 689 out of 1,000, against 657 for neobanks. Marcus ranked highest among high-yield savings providers with 739, ahead of Ally at 728. Awards, ratings and editorial reviews of this kind are, we infer, exactly the third-party evidence an assistant can cite and a cautious saver trusts.
Our inference on safety facts. Deposit insurance is where a chartered digital bank has a clear story, and where precision matters. The FDIC’s official sign states that each depositor is insured up to at least $250,000. The FDIC has also moved the compliance date for showing its official digital sign on websites and apps to January 1, 2027 (opens in a new tab), while it revises those rules. Banks that state the legal bank name behind each brand, the insurance facts and their conditions in plain text give assistants something accurate to repeat. Banks that bury them leave the answer to whatever page an assistant finds first.
What does GEO look like for a digital bank?
Generative engine optimization (GEO) keeps your rates, conditions and protections easy for assistants to find, check and repeat.
For a digital bank, the work usually covers:
- Rate pages written for checking. The current rate in text, the date it took effect, the conditions (minimums, bonus periods, caps) next to the number, and old promotional pages retired or updated.
- A clear entity. The legal bank name behind each brand, its regulator and its FDIC insurance facts on every product page, consistent everywhere they appear.
- Comparison-site accuracy. Correct, current listings on the rate tables and review sites that savers and assistants use. Many digital banks already pay these sites for referrals; the same listings now feed AI answers.
- Independent proof. Satisfaction rankings, editorial reviews and press coverage that confirm service, not only rate.
- Life-event content. Plain pages for savers with a lump sum, an emergency fund to build or a baby on the way, written without promising outcomes.
- Monitoring after every rate change. Ask assistants your savers’ questions after each change and correct wrong facts at their source, following how to fix wrong information about your brand in AI answers.
No one can guarantee that an assistant will name a bank. GEO makes sure that when it does, the rate, the conditions and the protections are right. For app-based finance more broadly, see how consumer fintech apps win customers when people ask AI about money. Vendors selling core, payments or fraud systems to banks face a different buyer; see how banking technology vendors reach bank shortlists.
Which questions about AI and deposits remain open?
Several: how often savers open accounts from AI answers, and how fast answers catch up with rate changes.
- No public link to funded accounts. We found no public data connecting AI answers to account openings or balances at any bank. The cross-industry evidence, thin as it is, is gathered in does AI visibility drive business results?.
- Rate accuracy is measured once, abroad. The Which? test was a single journalist’s check of one assistant in the UK. No comparable US test of savings rates has been published that we could find.
- Survey sources have interests. The Credit One survey was produced by a bank, and the UK figures come from a bank and a research agency. Treat them as signals.
- Our studies are snapshots. They cover US searches collected in September 2026; answers vary between assistants and over time.
Where should a digital bank begin?
Begin by asking assistants the questions your savers ask, then check every rate, condition and insurance fact in the answers.
A useful first review covers best-rate questions for your products, head-to-head comparisons with the banks you compete with, safety questions about online banks, and the switching questions that follow a rate change. It shows whether you are named, which banks and comparison sites appear instead, and whether your rates and protections are stated correctly.
If your deposit growth depends on savers choosing you when they move money, ask us to review how AI answers describe your bank. We will compare how assistants present you and your competitors, list the rates and facts they get wrong or miss, and plan the content, listings and coverage, reviewed with your compliance team, that keep those answers accurate after every rate change. Our generative engine optimization service page shows how that work is organized for a bank, from rate-page upkeep and listing corrections to plain pages for savers.
Frequently asked questions
Do people really ask ChatGPT where to open a savings account?
Many ask AI about savings. In J.D. Power’s survey, savings strategies were the top money topic people asked AI about (45%), and 59% used AI for banking and financial services occasionally. No public data counts account openings from AI answers.
Why would an AI answer show the wrong savings rate?
Rates change often, and answers can draw on old or promotional pages. In the Which? test, ChatGPT quoted a 4.5% rate for an account paying 3.75%.
Does being a chartered bank help in AI answers?
It gives you a clear safety fact to state. Whether an assistant repeats it depends on whether your pages say it plainly and consistently.
Can GEO make an assistant recommend our bank?
No. GEO makes accurate information about your bank easy to find and verify. It does not give personal financial advice, and its content should pass the same compliance review as your advertising.
Sources
- ABA Banking Journal (2025-09), Survey: Consumers increasingly turn to AI for financial advice (opens in a new tab)
- FDIC (2026-06), National Rates and Rate Caps, June 2026 (opens in a new tab)
- Ally Financial (2026-07-21), Second quarter 2026 press release
- Pulse 2.0 (2026-07), SoFi sees 43% revenue increase for Q2 2026 (opens in a new tab)
- Credit One Bank, via Stacker and KEYT (2026-07-15), The 4% flashpoint: Why Americans are quietly dumping traditional bank accounts (opens in a new tab)
- Which? (2026-03-26), Can AI really find you the best savings deal? (opens in a new tab)
- STRAT7 (2026), Who’s using AI for financial advice? New UK research (opens in a new tab)
- InvestmentNews (2024-03-27), Two-thirds of US savers are leaving interest on the table (opens in a new tab)
- J.D. Power, via 01net (2026-04-29), Online-only banking providers continue to win over customers with personalized digital experiences (opens in a new tab)
- FDIC (2025-12), Compliance date extension for sections 328.4 and 328.5 amendments (opens in a new tab)
- Federal Register, FDIC (2024-01-18), FDIC Official Signs and Advertising Requirements, False Advertising, Misrepresentation of Insured Status (opens in a new tab)
- Google Search Central (2025), AI features and your website (opens in a new tab)
- Underneath (2026), When does Google show an AI Overview? 1,248 US searches
- Underneath (2026), How faithfully do AI assistants quote software prices?
- Underneath (2026), How fresh are the pages AI engines cite?
- Underneath (2026), Same question, four countries: do AI recommendations change?
- Underneath (2026), Do Wikipedia and schema make AI assistants recommend a brand?